Ryohei Sasano
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Ryohei Sasano has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Hiroya Takamura
National Institute of Advanced Industrial Science and Technology
3 shared publications Manabu Okumura
2 shared publications- Manabu Okumura
Institute of Science Tokyo, Tokyo, Japan
2 shared publications - Akihiko Watanabe
Tokyo Institute of Technology
1 shared publication Daisuke Kawahara
Graduate School of Informatics Kyoto University
1 shared publication- Graham Neubig
Carnegie Mellon University, Pittsburgh, Pennsylvania, USA
1 shared publication
Latest Publications
- Open Access
Controlling Output Length in Neural Encoder-Decoders
Neural encoder-decoder models have shown great success in many sequence generation tasks. However, previous work has not investigated situations in which we would like to control the length of encoder-decoder outputs. This capability is crucial for applications such as text summarization, in which we have to generate concise summaries with a desired length. In this paper, we propose methods for controlling the output sequence length for neural encoder-decoder models: two decoding-based methods and two learning-based methods. Results show that our learning-based methods have the capability to control length without degrading summary quality in a summarization task.
Open AccessControlling Output Length in Neural Encoder-Decoders
Neural encoder-decoder models have shown great success in many sequence generation tasks. However, previous work has not investigated situations in which we would like to control the length of encoder-decoder outputs. This capability is crucial for applications such as text summarization, in which we have to generate concise summaries with a desired length. In this paper, we propose methods for controlling the output sequence length for neural encoder-decoder models: two decoding-based methods and two learning-based methods. Results show that our learning-based methods have the capability to control length without degrading summary quality in a summarization task.
- 20 February 2016
- Published by Springer Nature
- p. 108-119
Context Representation with Word Embeddings for WSD
- 20 February 2016
- Published by Springer Nature
- p. 108-119
Context Representation with Word Embeddings for WSD
- Open Access
- 1 January 2016
- Published by Japanese Society for Artificial Intelligence in Transactions of the Japanese Society for Artificial Intelligence
- Vol. 31 (5) , C-G41_1-G41_1
Generating Personalized Snippets forWeb Page Recommender Systems
Web page recommender systems usually provide users with titles and snippets of recommended pages when the systems present a list of recommendations. Snippets help users judge whether recommended web pages are relevant or not. However, while search engines usually show a text span around a search query as a snippet, web page recommender systems cannot leverage the snippet generation methods used by search engines because the recommender systems have no search queries. Web page recommender systems thus generally use lead sentences, i.e. the first sentences of web pages, as a snippet, but lead sentences are not necessarily relevant to user's interest. Furthermore since user's information needs can be different from each other, personalized snippets are desirable to support user's relevance judgment. Therefore, we propose a new method to generate personalized snippets for web page recommender systems that uses reasons why the web pages are recommended to the user. This use of reasons enables snippets to reflect the interest of the user. Furthermore, since our formulation does not depend on a certain recommender system, our method can be applied to diverse recommender systems. The experimental result on manually created dataset shows that our method is superior to the existing method and generic summarization model in terms of ROUGE-2. In addition, our method achieves comparable performance with the lead method despite that our method restricts itself to sentence selection while the lead method is free to extract a part of a sentence at the end of its snippets.
- 1 January 2016
- Published by Japanese Society for Artificial Intelligence in Transactions of the Japanese Society for Artificial Intelligence
- Vol. 31 (5) , C-G41_1-G41_1
Open AccessGenerating Personalized Snippets forWeb Page Recommender Systems
Web page recommender systems usually provide users with titles and snippets of recommended pages when the systems present a list of recommendations. Snippets help users judge whether recommended web pages are relevant or not. However, while search engines usually show a text span around a search query as a snippet, web page recommender systems cannot leverage the snippet generation methods used by search engines because the recommender systems have no search queries. Web page recommender systems thus generally use lead sentences, i.e. the first sentences of web pages, as a snippet, but lead sentences are not necessarily relevant to user's interest. Furthermore since user's information needs can be different from each other, personalized snippets are desirable to support user's relevance judgment. Therefore, we propose a new method to generate personalized snippets for web page recommender systems that uses reasons why the web pages are recommended to the user. This use of reasons enables snippets to reflect the interest of the user. Furthermore, since our formulation does not depend on a certain recommender system, our method can be applied to diverse recommender systems. The experimental result on manually created dataset shows that our method is superior to the existing method and generic summarization model in terms of ROUGE-2. In addition, our method achieves comparable performance with the lead method despite that our method restricts itself to sentence selection while the lead method is free to extract a part of a sentence at the end of its snippets.
- Open Access
- 1 January 2013
- Published by Association for Computational Linguistics (ACL)
- p. 1213-1223
Automatic Knowledge Acquisition for Case Alternation between the Passive and Active Voices in Japanese
Ryohei Sasano, Daisuke Kawahara, Sadao Kurohashi, Manabu Okumura. Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing. 2013.
- 1 January 2013
- Published by Association for Computational Linguistics (ACL)
- p. 1213-1223
Open AccessAutomatic Knowledge Acquisition for Case Alternation between the Passive and Active Voices in Japanese
Ryohei Sasano, Daisuke Kawahara, Sadao Kurohashi, Manabu Okumura. Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing. 2013.